All Forums > Team D v. Team efg% Debate
3/27/2013 7:31 PM (edited)

I’m sure some of the guys on here have advanced degrees in statistics and could provide more information than I am about to (I got an A in stats in undergrad, an A in a graduate level course, worked as a tutor for Econ and stats, and had an A- in my undergrad econometrics course) so I have some background but likely not as much as some people here.

I actually had not intended to share this, but maybe ash, dh or someone that has a bank of sim data saved could send me some files so I can produce more valid findings. I wish I had not been so lazy and had saved the sim data for the leagues I’ve played, but alas I did not, so the below just looks at the most recent PPL, which means that n=24 (below the central limit theorem of 30).

Basically what I did was run some correlations between different variables in the most recent PPL. Calculating the data was simple for all categories except team D (since the sim gives us the data in the team rankings section.) For team D I calculated a weighted average of total team D by taking the total number of minutes played for a team and dividing each player’s minutes in the sim by the team’s total, giving me the weight for each player, I then multiplied each player’s weight by that player’s D rating and added all of weighted averages together for a total team D.

Here are some of my findings:

The range for team D was 54.91 to 84.79 (the bottom three teams in D won 25, 28, and 30 games while the top three won 41, 49, and 57) however the correlation between team D and wins was only .4099 (ash had the 4th worst D and won 52 games while the team with the 2nd best D went 41-41).

The correlation between efg and wins was .5873. The range was .4713 (same team with 2nd best D that went 41-41) to .5437 (the same 57 win team (great team btw nc) that was also 3rd in team D).

The mean D for the league was 68.58 with a stdeva of 7.73 while the total efg for the league was .5137 with a team stdeva of .0191.

I also calculated a t-score for each team’s D and efg and combined the two to see how this value would correlate with W’s. The value here was .7296.

Again n=only 24, but efg seems to be a better predictor than D when it comes to W’s and looking at BOTH efg AND Def is a better predictor than efg alone.

I will say this for team D, the correlation between team D and opponent efg was -.8869. So, yea a great team D will impact how well your opponent shoots, but that in itself does NOT = W’s in the sim.

Really want to know how to win, outscore the other guy. The correlation between team point differential and W’s was a not so amazing .96997. (HOW to outscore the other guy is the ever burning question of course, because we KNOW it = wins.)

Oh and in case you are wondering rebounds and W’s had a .6387 correlation, which means by themselves they are more important that either D alone or efg alone.

Ash KNOWS EXACTLY what he’s talking about guys, heck I’d bet good money he has run his own models on the sim and has even better data than mine to produce more valid results. Final note is that team FT% and W’s had a .1953 correlation, i.e, worthless (at least for this set of 24) just as ash maintains.

3/26/2013 7:45 PM (edited)
I think the value of these things could vary wildly across cap levels. Definitely more data needed!

And I've had a bit more stats than you but I don't feel like checking your work! (Most of the stats I took was about proving why stuff works anyways! Though I do occasionally teach a 100 level stats course.)
3/26/2013 7:49 PM
Agreed logain and even though the PPL had a \$52 mill cap we had far fewer three's than a normal \$52mill draft league since we can't cherry pick the best 3pt seasons of all time. I'll be keeping everything moving forward, but it would be cool for someone to offer up some recent ODL or 52mill league data as well.
3/26/2013 7:55 PM
Open leagues would be good as well.
3/26/2013 8:05 PM
Logain, even though I'm only approaching about two thousand play-by-plays, I also believe there is a superstar "clutch" factor in the sim that is almost impossible to measure. With this current ODL I really considered going with a team efg% well north of 55% after I started with Lucas and Stockton, but I have seen a team's superstar win SO MANY games in crunch time that I decided I HAD to take Kobe and give back 2-3% of my team efg in the process. I also don't know EXACTLY where the WIS cutoff is for this, maybe 27.2% usage, maybe 30%, or maybe it's a top secret list of guys and usage does not matter, but players like Kobe, MJ, and LeBron win a heck of a lot more games at the end than they lose and way more than the non-superstars do.
3/26/2013 8:06 PM
The high rebounding-wins correlation makes me wonder if opponent efg is a better win predictor than team efg.
3/26/2013 8:15 PM
bargearse for this set of 24 opp efg and w's had a -.51396 correlation and efg differential had a .80433 correlation with W's. So for this data a team's own efg% was sliightly more important than their opponents, and winning the shooting battle is very important, but it clearly does not explain as much as winning the scoring battle.
3/26/2013 8:27 PM (edited)
Yeah makes sense of course. I just thought the rebounding correlation hinted at more defensive rebounds being available to the winning team, but that's really a function of both team's efg anyway I guess.

Not really possible to get the data, but it would be interesting to see if the numbers varied for fourth quarter performance too.

Don't know why I'm interested as I haven't played the basketball sim for a long time. I guess any stats get me thinking heh.
3/26/2013 9:23 PM
Very interesting analysis, nate.  You've actually run more numbers than I typically do, so I'm going to go at this from a different angle and then play the trump card.

My angle is the "why" it works the way it works.  First, a history lesson.

Used to, defense in the sim was virtually meaningless.  There was some effect on opponent shooting, but it was miniscule.  Also, it cost way more than it does now, so it just wasn't worth the added cost.  It was literally the last thing I (and many other owners) looked at when building a squad.  For years, we put up a fight with WIS to increase the effect of defense and/or decrease the cost of it so that it was worth it to maybe try and build a defensive squad.  The reason they gave us that it didn't seem to do anything actually made quite a bit of sense when examined in context with some other things.

The reason they said defense appeared to not do much was that all of the numbers used in the sim are based off real life statistics (ha - another rant for another time).  In real life, you didn't have these super star teams.  In a typical \$42m team right now, every person in your SL is almost always someone who has been an all-star multiple times or an extremely important role player on championship teams.  That makes a huge impact on everything... mostly on the offensive end.  You now are taking the most optimal seasons of everybody you can with the knowledge that these stats will create the outcome instead of knowing that the stats are a by product of the outcome.  So every owner pushes the envelope ever harder, tweaking this and tweaking that until you reach a point where no matter how good a defense faced, the offense is better.  Hard evidence later.

This is kind of consistent with real life.  Ever heard the term, "Good defense, better offense?"  Some guys just can't be guarded.  "Jordan's double teamed... no triple teamed... no whole team-teamed!"  This translates very well into the sim.

Look at how many factors can be used to pump up offensive (specifically shooting) efficiency:  fg%, ast%, tempo.

Defensive rating is the only factor that really can decrease shooting efficiency.  Worse, there's a broke mechanic (defensive positioning) that messes up things even worse (by creating foul issues).  Oh, and then there's another broken mechanic (double teaming) which has been tested time & again and shows virtually no difference in anything.  In fact, it seems to only enhance the guy double teamed (here comes the "yeah but I did it and it worked really well" sayings... guys... it's placebo.  You need a huge sample size with both the control & the variable to do accurate testing, and nobody really has the time, the money, or the desire to do the tests again).  Playing press D actually increases opponent shooting as well.

The point is that there are multiple things going into making someone shoot better.  The only thing working against that isn't bumped up by anything and is actually lessened by differing coach settings.

Let's go back in time again.

When we were in the fight to have defense be meaningful, the sim environment was vastly different.  These were the days of moses-worm-peja-allen-nash dominance.  Salaries were REALLY low and defensive ratings were even further out of whack than they are now.  You could build a better team for \$42 million (with a full 19,600 minutes) at that time than you can now for \$50 million.  So multiply the offensive prowess even more.  Oh yeah, and three point shooting got a double bonus from assist% and running up tempo (which stayed until very recently).

WIS finally gave in and changed many things that effect things as they are now.  For one, they went WAY up on salaries, specifically for high fg% and high usage%.  Then they backslid a bit and based it off of true shooting percentage instead of varying values for efg% and for ft%.  I explain why this is bad somewhere in the ODL XLII commentary thread.  Anyway... they made offense much more expensive to get while simultaneously drastically decreasing how much effect defense had on the salary formula.  That brought a lot of power back to defense.  Then they changed the way defensive ratings were determined (dropping rebounding percentage out of the equation - finally!)  For example, Troy Murphy was a 96 (which was max in the old system) defender... as was 100% of the top rebounders.  You essentially had a fantastic defense in the front court at all times.  That meant the only real difference seen between team to team defenses, before this change, was in the back court... and then you're talking about average team defenses of, for example, 90 vs 74.  Not much difference, especially in a sample size as small as 82 games.  So once they took the rebounding % out of defense, suddenly, there were many fewer good defenders, making aggregate defense much more noticeable.  However, since there are fewer good defenders, you'd expect the trend to show an increase in offensive efficiency.  It actually went the other way because of the offensive salary increases.

Then there was the 3pt%# glitch that jlg discovered which got fixed not too long ago.  When they fixed it, they also fixed the double bonus 3-point shooting was receiving.  I don't know if they did it on purpose or not, but they sure as shootin' fixed it.  So now, we're finally approaching a sim level where good defense and good offense are almost balanced (when it comes to cost).

But here's the trump card:

Say you're rocking a 100 defense, 50% efg% guard.  Say I'm rocking a 35 defense, 61% efg% guard.  Say all other factors influencing shooting are even.  Say that your 100 defense means my guard will lose 10% shooting effectiveness (it's no where near this much but for simplicity's sake we'll say it).  My guard still shoots 55%.

Say my 35 defense means your guard grants you an equal bonus (assuming 50 = no bonus, no penalty), which would be 3%.  Your guard shoots 51.5%

55% is still greater than 51.5%.

All possession factors considered even, I score more points than you, which means I get the W.  Does it happen every time?  No.  There's a high degree of randomness in this sim.  But I don't draft stats for a desired effect.  I draft stats to give me the best chance possible of achieving a desired effect.  That's a major difference in how I approach the sim and how many others approach it.

3/26/2013 9:31 PM
regarding rebounding:  I've seen that, in a variety of caps, my winningest teams tend to end up with higher than 33% orb & 75% drb (that's figured from results of the sim, not any aggregate totals from real life or otherwise).

Translated:

More than one out of every three of my missed shots leads to another shot attempt for me, while less than one out of every four of my opponent's missed shots leads to an extra shot for him.  That translates to extra possessions, which in coordination with the shooting struggle, leads to more consistent wins.

3/26/2013 11:37 PM
"I actually had not intended to share this, but maybe ash, dh or someone that has a bank of sim data saved could send me some files so I can produce more valid findings"

ehh flattered but i wouldnt group me in with ash or yourself....ive actually saved nothing over the years....dont really rely on data at all when drafting....to be honest....while i love stats and box scores i really only have a basic understanding of them(havent had a math class since high school)....what i do know is how players usually perform in the sim and what players go well together....not so different from a real nba team sometimes....for example take the 3 peat lakers with shaq...they had him and kobe...right away youre set with usage and you want those 2 shooting the ball almost every time....so you surround them with low usage guys like fisher,horry, fox etc...you also make sure they shoot the 3 so you have good floor balance since shaq owns the paint....simple stuff like that usually translates to the sim as well

of course since the sim changes you have to adapt as well....thats why ill be following along on this thread and paying very close attention to the odl results and the remaining dh52 season to see how much defense really matters

thats one of the main reasons i went for a half-court defensive team this time around.....ill get to see first hand if it was a huge mistake

keep an eye out on my division in the dh52(starting to like how that sounds) too....banditone built an unusual squad that kills teams on the boards and gets to the line a ton....great d too with all 5 starters playing 38+ a night with a 70+ d....but the offense is really crappy...rick friggin barry runs point and the team has no 3s....when i first saw the team i thought there was no way theyd compete for the division even with the boards and d....but 32 games in and whadaya know...banditone is in 1st with a team efg% under 50 while my team with an efg% close to 60 is 4th...banditone also plays press d...as does another team im trailing....im 3-5 against them....nate seems to also have recent success with the press....could be the recent changes also made press d an option now....or it just be too small a sample size and ill start crushing those teams with 140 point nights....really hope its the latter
3/26/2013 11:52 PM
better defense makes sense in the funk52... as if you don't spend the extra cash on the guys that play it, you'll end up with wasted salary/wasted minutes.  It's always been a struggle for me in the \$52m draft league to spend all of my salary.
3/26/2013 11:58 PM
Ash.. didnt they claim to reduce the chance of making a shot when fouled on a 3pa . Thus reducung 4pt plays. Could this be the cause of the drop you see in fg% of some bombers?
3/27/2013 1:32 AM
perhaps... but if fouled on a 3pa, the 3pa doesn't count unless you make it.
3/27/2013 7:34 AM
But if a player loses 20 4 pt plays.. that would be 20 less makes and 20 less attempts. Say a player was 100/250.. 40% on 3s. But 20 of those were 4 pt plays thst have now been reduced.. the player would then be 80/230 for 34.8%. Now this is probably an extreme example.. but 4 pt plays were very prevalent i think.
of 6
All Forums > Team D v. Team efg% Debate

Popular on WhatIfSports site: Baseball Simulation | College Basketball Game | College Football Game | Online Baseball Game | Hockey Simulation | NFL Picks | College Football Picks | Sports Games

© 1999-2013 WhatIfSports.com, Inc. All rights reserved. WhatIfSports is a trademark of WhatIfSports.com, Inc. SimLeague, SimMatchup and iSimNow are trademarks or registered trademarks of Electronic Arts, Inc. Used under license. The names of actual companies and products mentioned herein may be the trademarks of their respective owners.